Cold Email Follow-Up: A Nine-Day Plan, and When to Stop

Nobody replies to the first message. That is the ordinary case, not a verdict on your writing, and the whole value of outbound sits in what you send afterwards. Send nothing and you have bought a lottery ticket. Send ten and you have bought a spam complaint.

Below is a four-message plan for one prospect, running about nine days, with the wording and the reasoning for each. Ten minutes a day. Before the plan, two things that decide whether any of it matters: whether the mail arrives at all, and whether you are allowed to send it.

Before Any of This: Will the Message Arrive?

A sequence that lands in spam is a sequence that does not exist. Google published hard requirements for anyone reaching Gmail inboxes, and they have applied since 1 February 2024. Every sender needs SPF or DKIM on the sending domain, valid reverse DNS, a TLS connection and messages that follow the standard format. Anyone sending more than five thousand messages a day to Gmail addresses needs all three of SPF, DKIM and DMARC, plus one-click unsubscribe support, with the From domain aligned to the authenticated one.

The threshold that catches teams out is the complaint rate. Google asks senders to keep reported spam below 0.30% in Postmaster Tools, and separately advises aiming under 0.10% and never reaching 0.30% at all. Those are small numbers. A handful of annoyed recipients on a modest list will breach them, which is the real argument for stopping early rather than any theory about persistence.

On the legal side, if any recipient sits in the United States, the FTC's compliance guidance is short and worth reading in full. Headers and the From line must be accurate, the subject must reflect what is inside, a working postal address must appear, and an opt-out must function for at least thirty days and be honoured within ten business days. The guidance also prices failure: "Each separate email in violation of the CAN-SPAM Act is subject to penalties of up to $53,088", a figure updated for inflation in January 2024. Rules differ elsewhere, particularly across Europe, so check what applies where your recipients are and take proper advice rather than ours.

Four Messages on a Calendar

Day	What goes out	What it tests
1	Opening message, one reason and one ask	Whether the reason is specific enough to be real
2	The same point, re-shaped and re-timed	Whether format rather than content was the barrier
4	The ask alone, with a named time	Whether a cheaper reply unlocks an answer
8 or 9	A closing note that means it	Whether any interest exists at all

Day 1: One Reason, One Ask

Spend ten minutes finding something that is true about this company this month: a new role, a funding round, a job advert that implies a project, a post somebody wrote. Then write five lines or fewer.

Shape to copy: open with the observed fact and the date it happened. One sentence naming the problem that fact usually creates. One sentence on what you do about it, in the buyer's vocabulary rather than yours. Close with a single request carrying a specific day. No attachment, no second question, no postscript offering a deck.

What day one cannot tell you: why nobody answered. Plenty of recipients never saw the message, and the open rate will not settle the question either, which is the subject of our teardown of why cold email benchmarks disagree with each other.

Day 2: Same Point, Different Shape

Steli Efti's plan for Close, first published in 2023 and updated this year, is specific about the next move: send it at a different hour of the day, and "communicate the same message, just in a different format", long where you were short or short where you were long, without attachments and without inventing a new pitch.

Shape to copy: two lines. The outcome a comparable team got, stated as a number of weeks or a removed task rather than a percentage you cannot source. Then the same request as yesterday, unchanged.

What this cannot do: rescue a weak opening premise. If the first thirty prospects produce nothing from either message, the fault is in the list or the offer, not the cadence, and our piece on writing the customer profile as a checklist rather than a feeling is where to start over.

Day 4: The Ask Alone

Two days later, send the shortest message you will write all week. Close's instruction is to "succinctly restate your call to action" with no re-explanation at all.

Shape to copy: one sentence naming the topic, one naming a time and date. That is the entire message. A stated slot is cheaper to accept or refuse than an open invitation to find time, and refusal is a useful outcome.

What this cannot do: generate interest that was not there. It lowers the cost of replying, nothing more.

Day 8: Say You Are Stopping, Then Stop

Four or five days after the third message, close the loop. This is the last one.

Shape to copy: state plainly that you will stop writing, name the problem once more in a single clause, and leave a route back that costs them one reply. No deadline, no invented scarcity, no final offer.

Then do what you said. Remove the contact from the sequence, record the outcome, and move to the next account. A closing note you do not honour is the one message in the plan that reliably damages you.

Stop at Three Follow-Ups, Even When the Dashboard Says Keep Going

Here is the position, and it goes against the instinct every sequencing tool encourages. Hunter's analysis of 31 million cold emails sent by its users during 2025 found that a three-message sequence roughly doubled total replies against a single message, 6.8% against 3.3%, with the average across all sequences at 4.5% and the opening message alone producing 2.9%. Beyond three messages, the same analysis reports the average reply rate is lower rather than higher.

Read that carefully, because it is the part people skip. Most of the reply volume arrives on the first message, at 2.9% of the 4.5% average. The follow-ups are recovering a fraction, and each additional one is recovering a smaller fraction while adding the same complaint risk against a threshold measured in tenths of a per cent. Efti's own guidance wobbles between a ceiling of four and a recommendation to stop at three, which is itself a signal that the returns near the end are thin enough to argue about.

Your own data could overturn this by showing meetings, not replies, arriving from message five or later. If you can show that, extend the sequence and ignore us. Nobody has published it convincingly, and the figures above come from one vendor's customers, so the honest instruction is to test it on fifty prospects of your own and count booked conversations rather than anything an inbox reports back.

What Not to Send

• A fabricated reply thread or a subject line implying an earlier conversation. It costs trust immediately and sits badly against the accuracy requirements above.
• Daily messages. The compressed cadence reads as automation and earns complaints rather than replies.
• A ten-step sequence. The returns shrink and the domain reputation does not recover quickly.
• Attachments on a cold message. They slow delivery and nobody opens them from a stranger.

Where the Plan Fails

• The list is wrong. No cadence rescues the wrong recipient, and more follow-ups to a bad list is how complaint rates break.
• There was no reason to write. If you cannot name the specific thing that prompted the message, hold it. Our note on what personalisation actually means covers the difference between a reason and a detail.
• The account is worth more than an email. For a genuine target, add a call or a message on another channel around day three and shorten the writing further, as our guide to mixing pipeline channels sets out.

Four honest messages, re-shaped rather than repeated, finished by day nine. Count meetings per hundred contacted, never opens. No sequence can promise a reply, the benchmark figures here come from one provider's user base, and the only numbers that will ever govern your own plan are your reply rate and your complaint rate.